Forecasting food price inflation during global crises

In this paper, we consider the forecasting of domestic food price inflation (DFPI) using global indicators, with emphasis on episodes of macroeconomic turbulence, namely, the Global Financial Crisis (GFC) and the COVID-19 pandemic and its subsequent repercussions. Our monthly dataset covers about two decades for more than a hundred economies. We employ dynamic model averaging (DMA) to tackle both model uncertainty and parameter instability and produce pseudo out-of-sample forecasts. Thus, we are able to focus on the forecasting ability of the global predictors of DFPI before and during the global crises. We find evidence that the DMA specification tends to outperform statistical models frequently used in the literature such as random walks, autoregressive models, and time-varying parameter models, especially during global crises. We also identify the most successful predictors during the crises using their posterior probabilities of inclusion. By comparing the distributions of such probabilities, we find that the international food price inflation is the most useful predictor of DFPI for numerous countries during both crises. Other indicators such as domestic CPI inflation as well as the international inflation of agricultural commodities, fertilizers, and other food categories improved their forecasting ability, particularly during the COVID-19 period.